Model comparison
Grok 4.7 vs Grok Build 0.1: which xAI model to code with
Grok Build 0.1 costs $1.00 per cached coding session against $2.30 on Grok 4.7. Both follow xAI's caching and 200K rules; they differ on window and status.
· Prices as of September 28, 2026
Grok 4.7
xAI · Released September 21, 2026
xAI's top model for coding and knowledge work, which xAI says works longer on hard tasks and checks its own work more carefully.
Grok 4.7 facts and comparisonsGrok Build 0.1
xAI · Released May 2026 · Preview
xAI's dedicated agentic coding model, which also answers to the older grok-code-fast ids.
Grok Build 0.1 facts and comparisons
The short answer
Grok Build 0.1 costs less than Grok 4.7 on every example workload, $1.00 against $2.30 for the agentic coding session and $0.19 against $0.54 for output-heavy work, with input at half the price and output at a third. Grok 4.7 is xAI's current top model for coding and knowledge work, with a 500K window and a place in Cursor and GitHub Copilot. Grok Build 0.1 is xAI's early-access agentic coding model, with a 256K window, reachable through OpenRouter and OpenCode.
Choose Grok 4.7 if
- You want the model xAI calls its most capable for coding and knowledge work, rather than an early-access release.
- Some of your requests need more than 256K tokens, and Grok 4.7 accepts up to 500K.
- You work in Cursor or GitHub Copilot, which carry Grok 4.7 but not Grok Build 0.1.
- You value xAI's claim that Grok 4.7 works longer on difficult tasks and checks its own work more carefully.
Choose Grok Build 0.1 if
- Cost leads: $1 input and $2 output per million, against $2 and $6 on Grok 4.7.
- Your jobs produce a lot of output, where Grok 4.7 costs 2.8x as much on the example generation.
- Your scripts still call the older grok-code-fast ids, which now resolve to Grok Build 0.1.
- You want the model xAI trained specifically for agentic coding workflows.
Side by side
Specs and prices
| Fact | Grok 4.7 | Grok Build 0.1 |
|---|---|---|
| Maker | xAI | xAI |
| API model id | grok-4.7 | grok-build-0.1 |
| Released | September 21, 2026 | May 2026 |
| Status | Current | Preview |
| Context window | 500K tokens | 256K tokens |
| Max output | Not published | Not published |
| Open weights | No | No |
| Input, per 1M tokens | $2 | $1 |
| Cache hit, per 1M | $0.50 | $0.20 |
| Cache write, per 1M | $2 (same as input) | $1 (same as input) |
| Output, per 1M tokens | $6 | $2 |
| Runs in | Cursor, OpenCode, OpenRouter, and GitHub Copilot | OpenCode and OpenRouter |
Standard API rates in US dollars, as published by each maker on September 28, 2026. Batch and priority tiers, taxes, and subscription plans are not included. Grok 4.7: Once a prompt reaches 200K tokens, every token in the request costs $4 input, $1 cached, and $12 output per million. The US regional endpoint costs 10% more. Grok Build 0.1: Once a prompt reaches 200K tokens, every token in the request costs $2 input, $0.40 cached, and $4 output per million.
Cost
What the same work costs
The same token counts, priced at each model’s published rates. Cached tokens are billed at each maker’s cache prices, so the session shows what caching is worth on each model.
Real sessions differ: the two models count the same code as different numbers of tokens, and reasoning settings change how much each one writes. Your own history is the real test.
| Workload | Grok 4.7 | Grok Build 0.1 |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $2.30 | $1.00 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.36 | $0.17 |
| Output-heavy generation, 30K input, 80K output | $0.54 | $0.19 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $253.00 | $110.00 |
| Where the session’s cost goes | ||
| Cache writes | $0.80 | $0.40 |
| Cache reads | $1.00 | $0.40 |
| Uncached input | $0.20 | $0.10 |
| Output | $0.30 | $0.10 |
- caching saves on the session with Grok 4.7 (57%)
- $3.00
- caching saves on the session with Grok Build 0.1 (62%)
- $1.60
Two xAI models, one caching rulebook
Grok 4.7 and Grok Build 0.1 follow the same xAI caching rules. The API caches repeated prompt prefixes by itself, keeping the same conversation id across requests improves the hit rate, and writing the cache carries no fee. Only the price of a hit differs: $0.50 per million on Grok 4.7, 25% of its input price, and $0.20 on Grok Build 0.1, 20% of input.
They also share a long-context rule at the same line. Once a prompt reaches 200K tokens, every token in the request is billed at double: $4 input, $1 cached, and $12 output on Grok 4.7, and $2 input, $0.40 cached, and $4 output on Grok Build 0.1. Even on its long-context rates, Grok Build 0.1 matches Grok 4.7's standard input price and stays below its standard cached and output prices.
The windows differ: 500K on Grok 4.7 and 256K on Grok Build 0.1, and xAI publishes a maximum output for neither. On Grok Build 0.1 the long-context band is a narrow slice from 200K to the end of its window. On Grok 4.7 it covers more than half the window.
How much less Grok Build 0.1 costs
Input costs 2x as much on Grok 4.7 and output 3x. Both gaps show plainly in the uncached workloads: the large one-off review costs $0.36 against $0.17, and the output-heavy generation $0.54 against $0.19, 65% less on Grok Build 0.1.
The example agentic session costs $2.30 on Grok 4.7 and $1.00 on Grok Build 0.1, 2.3x apart. Cache reads are the largest line on Grok 4.7, $1.00 or 43% of its session, while on Grok Build 0.1 reads and writes tie at $0.40 each. Over 110 sessions a month the totals come to $253.00 against $110.00, a $143.00 difference.
Caching saves a larger share on Grok Build 0.1, 62% of its uncached session cost against 57% on Grok 4.7, because its hit is a smaller fraction of input. xAI's US regional endpoint also costs 10% more for Grok 4.7, according to its price note, and the tables use xAI's standard API prices.
Status, tools, and what xAI says each model is for
xAI released Grok 4.7 on September 21, 2026, and calls it "our most capable model for coding and knowledge work." It says Grok 4.7 works longer on difficult tasks, checks its own work more carefully, and rests on a new, larger base model trained with a longer reinforcement-learning run on many-hour tasks. Cursor's model list credits it to joint training by Cursor and xAI, and GitHub Copilot, OpenCode, and OpenRouter carry it too.
Grok Build 0.1 is older, announced as early access in May 2026. xAI describes it as "xAI's coding model, trained specifically for agentic coding workflows," with function calling, structured outputs, and reasoning, and it also answers to the older grok-code-fast ids. It is reachable through OpenRouter and OpenCode. GitHub Copilot retired Grok Code Fast 1 on May 15, 2026, and does not list Grok Build 0.1.
EveryToken prices both Grok models only when you use them through OpenRouter, from OpenRouter's catalog, which puts the two side by side in one history if you switch between them there.
Prompt caching
How xAI bills cached tokens
xAI
The xAI API caches repeated prompt prefixes automatically. Sending the same conversation id with each request raises the hit rate.
xAI lists no fee for writing the cache. A cache hit costs $0.50 per million tokens on Grok 4.7 and $0.20 on Grok Build 0.1.
Source: xAI docs: Prompt caching
Your own numbers
See what Grok 4.7 and Grok Build 0.1 really cost you.
everyaitoken reads your OpenRouter history on your Mac and prices every request at API rates, with what caching saved or cost. $9 once.
FAQ
Questions
Is Grok Build 0.1 cheaper than Grok 4.7?
It is cheaper on every rate and on all three example workloads. The agentic coding session costs $1.00 against $2.30, and output costs $2 per million against $6.
Is Grok Build 0.1 the same model as Grok Code Fast?
xAI says Grok Build 0.1 also answers to the older grok-code-fast ids, so requests to those ids reach it. Grok Code Fast 1 itself left GitHub Copilot on May 15, 2026.
Do both Grok models charge more above 200K tokens?
Yes. Once a prompt reaches 200K tokens, Grok 4.7 bills every token at $4 input, $1 cached, and $12 output per million, and Grok Build 0.1 at $2 input, $0.40 cached, and $4 output.
Which Grok model is available in Cursor?
Grok 4.7. Cursor lists it as trained jointly by Cursor and xAI, and also offers a faster Grok 4.7 Fast at double the rates. Cursor does not list Grok Build 0.1.